Journal of Chromatography A, 962 (2002) 95–103
www.elsevier.com / locate / chroma
High-performance liquid chromatography–mass spectrometry and
electron-capture dissociation tandem mass spectrometry of
osteocalcin
Determination of g-carboxyglutamic acid residues
a
b
b
a,
Hanna Niiranen , Bogdan A. Budnik , Roman A. Zubarev , Seppo Auriola *,
a
Seppo Lapinjoki
a
Department of Pharmaceutical Chemistry, University of Kuopio, P.O. Box 1627, FIN-70211 Kuopio, Finland
b
Department of Chemistry, University of Southern Denmark, DK-5230 Odense M, Denmark
Received 11 January 2002; received in revised form 18 April 2002; accepted 18 April 2002
Abstract
Two mass spectrometry methods, high-performance liquid chromatography combined on-line with electrospray ionization
mass spectrometry (HPLC–ESI-MS) and electron-capture (EC) dissociation tandem mass spectrometry (MS–MS), were
applied for structural analysis of bovine and human osteocalcins. Osteocalcin contains g-carboxyglutamic acid (Gla)
residues, which bind metal ions, among its amino acids. Ethylenediaminetetraacetic acid (EDTA) was added to all samples in
order to chelate bound metal ions. After elimination of interfering metal ions MS spectra became uncomplicated to interpret.
EDTA is incompatible with ESI and it was removed from samples using either on-line HPLC or micropurification method.
The number of Gla residues varies in osteocalcin. These subforms, which contain different amounts of Gla residues, were
separated using the HPLC–ESI-MS method. In order to determine locations of Gla residues in human osteocalcin, which
contained two Gla residues, dissociation MS–MS method was successfully applied.  2002 Published by Elsevier Science
B.V.
Keywords: Electron-capture dissociation; Mass spectrometry; Carboxyglutamic acid; Osteocalcin
1. Introduction
Osteocalcin (OCN), also called bone Gla protein,
is a bone matrix protein of 49 amino acid residues in
humans [1]. OCN was independently discovered by
two separate research groups of Hauschka et al. [2]
and Price et al. [3]. OCN is presently known as the
most abundant non-collagenous bone protein [4,5].
*Corresponding author. Fax: 1358-17-162-456.
E-mail address: [email protected] (S. Auriola).
OCN is secreted by osteoblasts and is a specific
biochemical marker for bone formation [6]. A special feature of OCN is its g-carboxyglutamic acid
(Gla) residues, which are located at positions 17, 21
and 24 [7]. The Gla residues increase the affinity of
OCN to calcium [8]. In the presence of calcium the
polypeptide backbone of OCN adopts a a-helical
conformation [9]. It is demonstrated that the Gla
residue at position 17 is essential for calcium-dependent conformational transition of OCN [10]. The
space between the Gla residues in three dimensional
0021-9673 / 02 / $ – see front matter  2002 Published by Elsevier Science B.V.
PII: S0021-9673( 02 )00451-X
96
H. Niiranen et al. / J. Chromatogr. A 962 (2002) 95–103
structure correspond to the interatomic distance of
calcium atoms in hydroxyapatite lattice and it is
demonstrated that this interaction accounts for tight
adsorption of OCN onto hydroxyapatite surface and
OCN’s affinity for bone [8,9].
Mass spectrometry plays an important role in the
characterization of peptides and proteins [11]. Detection of ever more unstable modifications, such as
glycosylation, carboxylation, oxidation, and phosphorylation, to biomolecules is possible using methods as electrospray ionization mass spectrometry
(ESI-MS) [12]. However, further microcharacterization of these unstable species with such energetic
methods can be difficult as a result of ejection of the
modification before backbone bond cleavage [12].
Formation of g-carboxyglutamic acid residue from
glutamic acid is an example of labile post-translational modifications. Several mass spectrometric
methods have been introduced to analyze Gla residues: gas chromatography–mass spectrometry (GC–
MS) [13–15], matrix-assisted laser desorption ionization mass spectrometry (MALDI-MS) [1,16,17].
However, these methods either need reasonably
complicated sample preparation methods or spectra
give inadequate data, when further analyzing methods are needed to identify the sample. Electroncapture (EC) dissociation is a new tandem MS (MS–
MS) method [18,19]. EC dissociation cleaves peptide
backbones primarily at the C a –N bond rather than at
the amide linkage as with collisionally activated
dissociation (CAD). The potential of EC dissociation
for analysis and sequencing of post-translationally
modified peptides and proteins has been demonstrated for O-glycosylated peptides [20], sulfapeptides [12,21], phosphorylated peptides [22], phosphorylated proteins [23] and g-carboxylated peptides
[12]. As a gentle fragmentation method EC dissociation does not induce loss of labile conjugated groups
in peptides in MS–MS experiments. A technical
description for EC dissociation will not be reviewed
here, as this has been covered in detail recently [24].
In this study the aim was to develop a MS method
to analyze OCN and its Gla residues. First a highperformance liquid chromatography–electrospray
ionization mass spectrometry (HPLC–ESI-MS)
method was successfully developed to separate subforms of OCN containing different amounts of Gla
residues. Further microcharacterization of OCN was
impossible with this HPLC–ESI-MS method: CO 2 of
Gla was ejected by ESI-MS–MS during CAD. The
ejection of CO 2 occurred before backbone bond
cleavage. Secondly an EC dissociation MS–MS
method was applied in order to locate Gla residues in
OCN peptide. The ECD-MS–MS method has been
demonstrated to be suitable for analysis of small 28
residue peptides containing Gla residues [12], however suitability for larger peptides containing Gla
residues has not been demonstrated earlier. Here we
report application of EC dissociation MS–MS for
determination of g-carboxylation sites in OCN.
Using these two methods it is possible to fully
characterize OCN and its subforms, which contain
different amounts of Gla residues. These methods
provide a useful tool to determine OCN’s exact
function in bone and to evaluate its importance as a
biochemical marker of bone function. These methods
are also useful in developing more stable peptide
drug formulations because it is possible to get more
information on labile post-translational modifications
of ever more larger proteins.
2. Experimental
2.1. Chemicals
Bovine osteocalcin (bOCN) was isolated from
bovine bone powder using a previously described
method [25] and further purified by a method
described by Kaartinen et al. [26]. The amino acid
sequence (Fig. 1) contains two to three residues of
Gla at positions 17, 21 and 24, hydroxyproline (Hyp)
at position 9 and disulfide bond between cysteine
residues (Cys-23 and Cys-29) [9]. Samples were
diluted with water to a final concentration. (Glu-17,
Gla-21, Gla-24)–Osteocalcin human 1–49 (hOCN)
was obtained from Peninsula Labs. (Belmont, CA,
USA). The amino acid sequence contains two Gla
residues at positions 21 and 24 and disulfide bond
between cysteine residues (Fig. 1). Solid protein was
dissolved in concentration 50 mg / ml in phosphate
buffer pH 7 obtained from FF-Chemicals (Yli-Ii,
Finland) [27], divided into aliquots and stored in a
freezer (220 8C).
Ethylenediaminetetraacetic acid (EDTA) Titriplex
III and formic acid were obtained from Merck
H. Niiranen et al. / J. Chromatogr. A 962 (2002) 95–103
97
Fig. 1. The primary structure of bovine osteocalcin 1–49 and of human osteocalcin 1–49 (Glu-17, Gla-21, 24). P*5hydroxyproline,
E5g-carboxyglutamic acid.
(Darmstadt, Germany), dithiothreitol was from
Boehringer–Mannheim (Germany), urea was from
Riedel-de Haen (Seelze, Germany), ammonium
hydrogencarbonate was from Fluka (Buchs, Switzerland), acetonitrile from Rathburn (Walkerburn, UK)
and Poros R2 from PerSeptive Biosystems (Cambridge, MA, USA). Water was deionized using a
Milli-Q system from Millipore (Bedford, MA, USA).
2.2. Methods
2.2.1. Chelation of metal ions with EDTA
bOCN was analyzed both without any additional
EDTA and with EDTA. EDTA was dissolved in
water and added to samples in order to get a final
concentration from 5 to 10 mM.
2.2.2. Determination of Gla residues
In order to determine the Gla residues the disulfide
bond was reduced and the Cys residues were
alkylated using the following procedure: 30 ml
bOCN (20 mg / ml) was taken and 3.5 ml EDTA (100
mM), 25 ml NH 4 HCO 3 (0.1 M) and 6 ml dithiothreitol (100 mM DTT in 0.1 M NH 4 HCO 3 ) was
added and stirred carefully. The solution was incubated at 56 8C for 15 min. After incubation 4 ml of
0.5 M iodoacetamide in 0.1 M NH 4 HCO 3 was added
and the solution was protected from light and
incubated at room temperature for 15 min.
2.2.3. HPLC–ESI-MS
2.2.3.1. HPLC–ESI-MS Method A
Positive ion mass spectra were acquired with online HPLC–ESI-MS. The measurements were carried
out using a LCQ quadrupole ion trap mass spectrometer equipped with an ESI ion source (Finnigan
MAT, San Jose, CA, USA), a Rheos 4000 HPLC
pump (Flux Instruments, Danderyd, Sweden) and a
Rheodyne 7725 injector with a 20 ml loop (Cotati,
CA, USA). The spray was stabilized using nitrogen
sheath gas flow (value 80–85). The parameters used
in the instrument were optimized using the auto-tune
method. Tuning was done using a MRFA (5Met–
Arg–Phe–Ala) standard. The column was a reversedphase HPLC column (Syncropak RP-8, 5032.1 mm,
SynChrom I, Lafayette, IN, USA). The gradient was
2 to 70% acetonitrile (containing 50 mM HCOOH)
in 40 min, flow 200 ml / min. Samples were injected
using LaChrom Autosampler L-7200 from Merck–
Hitachi (Hitachi, Tokyo, Japan), injection volume 20
ml. The full scan mass spectra from m /z 410 to m /z
2000 were measured.
2.2.3.2. HPLC–ESI-MS Method B
Positive ion mass spectra were acquired with online micro (m) HPLC–ESI-MS. The HPLC–ESI-MS
instrument described in Method A was used with the
following modifications: The flow from the HPLC
pump was split with a flow controller (Acurate, LC
Packings, Netherlands). The column was a Vydac C 8 ,
˚ (LC Packings). The
15030.3 mm (5 mm, 300 A)
gradient was 7 to 70% acetonitrile (containing 50
mM HCOOH) in 30 min, flow 4 ml / min. The
original ESI orifice was replaced by a PicoTip
adapter with fused-silica needles (New Objective,
Cambridge, MA, USA).
2.2.4. EC dissociation MS–MS
In order to chelate metal ions EDTA was added to
the hOCN sample. EDTA has to be removed before
MS analysis and the following procedure was used:
Micro column of Poros R2 was prepared in a gelloader tip using Poros R2 suspension in 1:1 methanol
and water. The hOCN solution containing 10 mM
EDTA was loaded into the column, followed by a
98
H. Niiranen et al. / J. Chromatogr. A 962 (2002) 95–103
washing with 5% formic acid. Peptides were eluted
using 50% methanol in 5% formic acid and loaded
into the MS sample capillary.
A 4.7 Tesla Ultima (IonSpec, Irvine, CA, USA)
Fourier Transform Ion Cyclotron Resonance (FTICR) mass spectrometer was used to perform electron-capture dissociation. External accumulation in
the hexapole of the ESI-source (Analytica of Branford, MA, USA, modified with a heated metal
capillary), followed by gated trapping was used
before the desired charge states were selected by
applying a pre-programmed waveform. Isolated cations were irradiated by electrons from a heated
tungsten filament for 12 s. Due to the electron
thermal energy and the non-zero potential in the
center of the cell, 11.25 V bias of the filament
center was used to ensure ,0.2 eV electron energy in
the region occupied by the parent ions.
3. Results and discussion
3.1. Chelation of metal ions with EDTA
bOCN was analyzed without any additional EDTA
(Fig. 2A) using the HPLC–ESI-MS Method A. The
HPLC–ESI-MS base peak ion chromatogram shows
several peaks at retention time 17–18 min. The
molecular ions (M13H)31 at m /z 1950.7, (M1
4H)41 at m /z 1463.7, (M15H)51 at m /z 1171.2 and
(M16H)61 at m /z 976.0 give an average mass of
5850.2 (Fig. 2A, Table 1). The expected average
molecular mass (M Ex
r ) for bOCN, including three Gla
and one Hyp residues and a single disulfide bond, is
5850.3. The molecular ions (M13H)31 at m /z
1936.4, (M14H)41 at m /z 1452.6, (M15H)51 at
m /z 1162.3 and (M16H)61 at m /z 968.8 give an
average mass of 5806.5 (Fig. 2A, Table 1). M Ex
for
r
bOCN, including two Gla and one Hyp residues and
a single disulfide bond, is 5806.3. These molecular
ions correspond to the expected molecular masses,
but the HPLC–ESI-MS spectra show also additional
molecular ions, which differ from the expected ions.
These molecular ions give average molecular masses
of, for example, 5873.8, 5903.4 and 5906.9 (Table
1). The difference between the expected and the
observed is 23.5, 53.1 or 56.6 units. These differences can be explained by the affinity of metal ions,
for example, sodium (atomic mass 22.99), magnesium (atomic mass 24.31), chromium (atomic mass
52.00) or iron (atomic mass 55.85) to the OCN
molecule. After this observation OCN was analyzed
with the addition of EDTA (Fig. 2B). The HPLC–
ESI-MS base peak ion chromatogram shows one
major peak. This peak corresponds both to molecular
ions from bOCN including three Gla residues (triGla bOCN) and to molecular ions from bOCN
including two Gla residues (di-Gla bOCN). The two
forms of bOCN co-elute and are therefore found in
the same peak. Any interfering molecular ions,
which include metal ions, are not observed because
of the addition of EDTA.
Metal ions interfere with the analysis, but they
have a remarkable effect on the stability of OCN. A
sample containing bOCN 1 mg / ml in 0.1 mM EDTA
was stored at 23 8C for 24 h. MS spectra of degradation fragments showed that only 12.5% of the
original peptide was left. Major degradation products
were fragments 15–49 (56.5%) and 1–14 (7%).
Moreover, dimers formed via the re-organization of
disulfide bonds were found in the spectra (24%). For
that reason EDTA was added only to those samples,
which were going to be analyzed immediately.
The majority of the components of the peptide
mixture were successfully separated by HPLC. However, those OCN subforms, which contain either two
or three Gla residues and thus differ one Gla from
each other, did co-elute. These forms were separated
in the mass spectra according to their masses. Buffer
salts, for example EDTA, which was added to OCN
samples, may interfere with the electrospray. For that
reason it is a convenient way to use on-line LC in
order to remove salts.
3.2. Determination of Gla residues
Further analyzing of bOCN was carried out using
HPLC–ESI-MS Method B. EDTA was added to all
samples. According to our previous observations
(Fig. 2B) the degree of carboxylation varies: from
mass spectra we obtain molecular masses for both
three and two Gla residues including forms of
bOCN. In the HPLC the forms co-elute, but they can
be separated in the mass spectra according to their
masses. CAD-MS–MS spectra were obtained from
molecular ions (M15H)51 at m /z 1171.0 (tri-Gla
H. Niiranen et al. / J. Chromatogr. A 962 (2002) 95–103
99
Fig. 2. HPLC–ESI-MS base peak ion chromatogram of bovine osteocalcin (20 mg / ml, 20 ml) without EDTA (A) and with 10 mM EDTA
(B). Analysis conditions are described in Method A. (M15H)51 and (M16H)61 (text in box) are molecular ions from osteocalcin, which
has two Gla residues, other molecular ions are from osteocalcin, which has three Gla-residues.
H. Niiranen et al. / J. Chromatogr. A 962 (2002) 95–103
100
Table 1
Obtained and expected molecular masses of the bovine osteocalcin 1–49 a
Molecular mass
expected
(M rEx )
Gla residues
Molecular mass
observed
(M Obs
)
r
Ions observed
DMr
Obs
(M Ex
)
r 2M r
5850.3
3
5850.2
(M13H)31
(M14H)41
(M15H)51
(M16H)61
1950.7
1463.7
1171.2
976.0
0.1
5806.3
2
5806.5
(M13H)31
(M14H)41
(M15H)51
(M16H)61
1936.4
1452.6
1162.3
968.8
0.2
5850.3
3
5873.8
(M13H)31
(M14H)41
(M15H)51
(M16H)61
1958.8
1469.4
1175.8
980.0
23.5
Metal ion
5850.3
3
5903.4
(M13H)31
(M14H)41
(M15H)51
(M16H)61
1969.3
1477.1
1181.5
984.6
53.1
Metal ion
5850.3
3
5906.9
(M13H)31
(M14H)41
(M15H)51
(M16H)61
1970.2
1478.4
1182.4
984.9
56.6
Metal ion
a
Comment
Gla-residues5g-carboxyglutamic acid residues. Used molecular masses are average.
bOCN) and (M15H)51 at m /z 1162.2 (di-Gla
bOCN) (Fig. 3). The two spectra obtained were
almost identical: one shows molecular ion (M1
5H)51 at m /z 1144.3 and the other shows (M15H)51
at m /z 1144.6. In order to eliminate the disulfide
bond the bOCN was reduced and Cys residues were
carbamidomethylated (CAM). In the mass spectra of
modified bOCN (mod-bOCN) there are molecular
ions (M13H)31 at m /z 1989.3, (M14H)41 at m /z
1492.1, (M15H)51 at m /z 1194.2 and (M16H)61 at
m /z 995.2 and the observed average molecular mass
is 5965.1. The M Ex
of tri-Gla mod-bOCN is 5966.3.
r
The obtained mass for di-Gla mod-bOCN was
5922.4 and the M Ex
5921.3. CAD-MS–MS spectra
r
from (M15H)51 at m /z 1194.2 and (M15H)51 at
m /z 1186.0 were almost identical again (Fig. 4): one
shows (M15H)51 at m /z 1167.7 and the other shows
(M15H)51 at m /z 1167.8. The explanation for the
peaks 1144.3 and 1144.6 (unmodified bOCN) and
1167.7 and 1167.8 (CAM-modified bOCN) in MS–
MS spectra is: Gla contains a labile carboxy group,
which is ejected during CAD. Molecular masses
obtained from the molecular ions in MS–MS spectra
are 5716.5 (1144.3 51 ) and 5718.0 (1144.6 51 ) and
5833.5 (1167.7 51 ) and 5834.0 (1167.8 51 ). When
these masses are compared to di-Gla and tri-Gla
bOCN (both unmodified and CAM-modified), we
can see a difference of approximately 132.6 or 88.2
units. The number 132.6 correlates to loss of three
carboxy groups and 88.2 correlates to loss of two
carboxy groups. Multiple MS (MS n) can be applied
for further characterization of the peptide, but obtained daughter ions show only Glu residues since
the g-carboxy groups are ejected during the first
CAD.
3.3. EC dissociation MS–MS
Fourier transform (FT) EC dissociation MS–MS
mass spectra (Fig. 5) show molecular ions and
fragment ions of Glu-17, Gla-21, Gla-24–hOCN 1–
49. Totally, 40 scans were integrated. The spectrum
H. Niiranen et al. / J. Chromatogr. A 962 (2002) 95–103
101
Fig. 3. HPLC–ESI-MS base peak ion chromatogram of bovine osteocalcin (8.7 mg / ml, 5 mM EDTA). Analysis conditions are described in
HPLC–ESI-MS Method B. (M16H)61 and (M15H)51 (text in box) are molecular ions from osteocalcin with two Gla residues, other
molecular ions are from osteocalcin with three Gla residues.
in Fig. 5 contains preferentially c and z ions characteristic for EC dissociation; some amount of b, y
fragmentation is also present, originating from collisional excitation during the charge-state isolation. No
backbone fragmentation is observed from the region
protected by the internal disulfide bond (residues
23–29). Outside the ring, all inter-residue bonds
were cleaved except those on the N-side of the Pro
residue that are EC dissociation-immune [18]. No
losses are observed from the even-electron c-ions;
minor CO 2 losses were detected from the radical
z-fragments, consistent with their lower stability
[20]. Since the ring contained only one glutamic acid
residue, and the total number of g-carboxylated
residues was known from the molecular mass shift,
the observed cleavages allowed for unequivocal
assignment of the modification to Glu-21 and Glu24. The signal for the uncarboxylated form is less
than 5% and for that reason Glu-17 and Glu-31 were
not to be found modified. Therefore, a single EC
dissociation spectrum was sufficient for complete
characterization of the g-carboxylation pattern of
hOCN. At this moment it is not possible to connect
on-line HPLC to EC dissociation MS–MS equipment
and thus the results are not in chromatographic scale.
Further study is needed to develop an on-line LC–
EC dissociation MS–MS method.
4. Conclusion
The interfering effect of metal ions to MS spectra
was eliminated using an addition of EDTA to the
osteocalcin sample. Osteocalcin subforms, which
contain different amounts of Gla residues among its
amino acids, were separated using the applied on-
102
H. Niiranen et al. / J. Chromatogr. A 962 (2002) 95–103
Fig. 4. MS–MS spectra of ion (M15H)51 at m /z 1186.0 and ion (M15H)51 at m /z 1194.2.
line HPLC–ESI-MS method. Localization of Gla
residues of the peptide was done using the applied
EC dissociation MS–MS method. According to this
study these MS methods may help to determine the
exact function of OCN in bone. When developing
more stable peptide drug formulations these MS
methods are suitable for qualitative analysis.
Acknowledgements
Financial support from the Technology Development Centre of Finland (TEKES, Finland) is gratefully acknowledged. The Odense group is grateful to
P.B. Hansen and T. Jensen for technical assistance;
the work was supported by the grants STVF0001242 and SNF-51-00-0358.
References
Fig. 5. Electron-capture dissociation MS–MS spectrum (40 scans)
of Glu-17, Gla-21, Gla-24–human osteocalcin 1–49.
¨ ¨
[1] J. Hellman, S.-M. Kakonen,
M.-T. Matikainen, M. Karp, T.
¨
¨¨ ¨
Lovgren,
H.K. Vaananen,
J. Bone Miner. Res. 11 (1996)
1165.
[2] P.V. Hauschka, J.B. Lian, P.M. Gallop, Proc. Natl. Acad. Sci.
USA 72 (1975) 3925.
H. Niiranen et al. / J. Chromatogr. A 962 (2002) 95–103
[3] P.A. Price, J.W. Poser, N. Raman, Proc. Natl. Acad. Sci. USA
73 (1976) 3374.
[4] P.A. Price, M.K. Williamson, J.W. Lothringer, J. Biol. Chem.
256 (1981) 12760.
[5] P.A. Price, M.K. Williamson, J. Biol. Chem. 256 (1981)
12754.
[6] E.D. Daniels, J.M. Pettifor, G.P. Moodley, Eur. J. Pediatr.
159 (2000) 730.
[7] J.W. Poser, F.S. Each, N.C. Ling, P.A. Price, J. Biol. Chem.
255 (1980) 8685.
[8] P.V. Hauschka, J.B. Lian, D.E.C. Cole, C.M. Gundberg,
Physiol. Rev. 69 (1989) 990.
[9] P.V. Hauschka, S.A. Carr, Biochemistry 21 (1982) 2538.
[10] M. Nakao, Y. Nishiuchi, M. Nakata, T. Kimura, S. Sakakibara, Pept. Res. 4 (1994) 171.
¨
[11] A.F.R. Huhmer
et al., Anal. Chem. 69 (1997) 29R.
[12] N.L. Kelleher et al., Anal. Chem. 71 (1999) 4250.
[13] S.A. Carr, K. Biemann, Biomed. Mass Spectrom. 7 (1980)
172.
[14] S.A. Carr, P.V. Hauschka, K. Biemann, J. Biol. Chem. 256
(1981) 9944.
[15] K. Rose, J.D. Priddle, R.E. Offord, M.P. Esnouf, Biochem. J.
187 (1980) 239.
103
[16] T. Matikainen et al., J. Bone Miner. Res. 14 (1999) 431.
¨ ¨
¨
[17] S.-M. Kakonen,
J. Hellman, K. Pettersson, T. Lovgren,
M.
Karp, Protein Expr. Purif. 8 (1996) 137.
[18] R.A. Zubarev, N.L. Kelleher, F.W. McLafferty, J. Am.
Chem. Soc. 120 (1998) 3265.
[19] R.A. Zubarev et al., J. Am. Chem. Soc. 121 (1999) 2857.
[20] E. Mirgorodskaya, P. Roepstorff, R.A. Zubarev, Anal. Chem.
71 (1999) 4431.
[21] K.F. Haselmann et al., Anal. Chem. 73 (2001) 2998.
[22] A. Stensballe, O.N. Olsen, J.V. Olsen, K.F. Haselmann, R.A.
Zubarev, Rapid Commun. Mass Spectrom. 14 (2000) 1793.
[23] S.D.-H. Shi, M.E. Hemling, S.A. Carr, D.M. Horn, I. Lindh,
F.W. McLafferty, Anal. Chem. 73 (2000) 19.
[24] R.A. Zubarev et al., Anal. Chem. 72 (2000) 563.
[25] G. Colombo, P. Fanti, C. Yao, H.H. Malluche, J. Bone Miner.
Res. 8 (1993) 733.
[26] M.T. Kaartinen, A. Pirhonen, A. Linnala-Kankkunen, P.H.
¨
¨ ¨ J. Biol. Chem. 272 (1997) 22736.
Maenpaa,
¨ ¨
[27] H. Vahatalo,
S. Auriola, S. Lapinjoki, J. Chromatogr. A 846
(1998) 49.